We wrote 100,000 lines of Swift entirely with AI — here's what building Solar Explorer 3D taught us
Hey Product Hunt! 👋
We're launching Solar Explorer 3D this Friday — a physics-accurate, interactive 3D solar system simulator for iOS. Before launch day, I wanted to share something about how we built it, and open a conversation with the maker community.
Solar Explorer 3D is our first project where every single line of code was AI-generated. 100,000+ lines of Swift, written entirely with Droid, Claude Opus, and Codex — no manually written code.
This wasn't a limitation. It was a deliberate experiment: how far can AI-assisted development go on a genuinely complex, production-grade app?
Here's what's inside:
3D real-time visualization of 50+ celestial bodies
Physics-accurate orbital mechanics powered by NASA JPL Horizons ephemeris data
Mission replay for Voyager 1 & 2, New Horizons, Pioneer 10 & 11, and more (new in v1.5.0)
Time controls, landing perspectives, multilingual support (10 languages)
Things we'd love to hear from this community:
For experienced developers: When you already know how to code, does AI-generated code feel like leverage or like liability? Where does the human judgment become more critical, not less?
For anyone working on complex AI workflows: Droid's long-context handling was essential for us — how are you managing context and coherence on large codebases?
For indie makers: Is 100k lines of AI-generated code a milestone, or just the new baseline? What does "software craftsmanship" mean when the machine writes it all?
Pre-launch page: https://www.producthunt.com/products/solar-explorer-3d/solar-explorer-3d/prelaunch
Happy to go deep on the technical side — architecture decisions, debugging AI-generated code at scale, or anything else. Let's talk.
Replies